fix prev link in docs

Hi,

On Thu, 11 Jun 2026 at 00:46, Nathan Bossart <nathandbossart@gmail.com>
wrote:

Right now, if you go to the following page and click "Prev", it'll send you
to the middle of the previous chapter:

https://www.postgresql.org/docs/devel/bgworker.html

My research indicates that this might caused by a bug in libxml2, but
regardless, a minimal fix is to change the link generation code to use
[position()=last()] instead of [last()] in the predicate on the union of
reverse axes. Patch attached.

Thanks for the patch.

I tested it and with patch the Prev tab works fine towards 45.6.
(previous chapter's end)

Regards,
Ayush

On Sat, Jun 13, 2026 at 02:14:26AM +0530, Ayush Tiwari wrote:

Thanks for the patch.

I tested it and with patch the Prev tab works fine towards 45.6.
(previous chapter's end)

Thanks for reviewing. Committed.

--
nathan

Nathan Bossart <nathandbossart@gmail.com> writes:

On Sat, Jun 13, 2026 at 02:14:26AM +0530, Ayush Tiwari wrote:

I tested it and with patch the Prev tab works fine towards 45.6.
(previous chapter's end)

Thanks for reviewing. Committed.

I noticed that building the HTML docs got really painfully slower
recently, and the reason seems to be this patch. If you look
at guaibasaurus' runtimes for instance:

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=guaibasaurus&amp;dt=2026-06-15%2012%3A20%3A02

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=guaibasaurus&amp;dt=2026-06-15%2020%3A20%3A02

the time for the "make-doc" step went from 00:46 right before
this patch to 3:39 right afterward. That's consistent with
what I'm seeing locally. I don't find that acceptable.

(Oddly, the time to make PDFs hasn't increased, per BF member
alabio.)

regards, tom lane

On Wed, Jun 17, 2026 at 01:40:07PM -0400, Tom Lane wrote:

I noticed that building the HTML docs got really painfully slower
recently, and the reason seems to be this patch. If you look
at guaibasaurus' runtimes for instance:

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=guaibasaurus&amp;dt=2026-06-15%2012%3A20%3A02

https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=guaibasaurus&amp;dt=2026-06-15%2020%3A20%3A02

the time for the "make-doc" step went from 00:46 right before
this patch to 3:39 right afterward. That's consistent with
what I'm seeing locally. I don't find that acceptable.

Oh, interesting. I somehow hadn't noticed that in my testing. I had a
more invasive form of the patch that I think might improve matters.
Double-checking...

--
nathan

On Wed, Jun 17, 2026 at 12:47:04PM -0500, Nathan Bossart wrote:

Oh, interesting. I somehow hadn't noticed that in my testing. I had a
more invasive form of the patch that I think might improve matters.
Double-checking...

The attached should resolve the issue without regressing build time.

--
nathan

Nathan Bossart <nathandbossart@gmail.com> writes:

The attached should resolve the issue without regressing build time.

Fixes the build-time problem for me, thanks!

(I didn't look at the output, except for the head page.)

regards, tom lane

On Wed, Jun 17, 2026 at 03:25:41PM -0400, Tom Lane wrote:

Fixes the build-time problem for me, thanks!

Committed.

--
nathan